Hi Philippe, there is no option to exclude atoms from refinement in phenix.rea_space_refine, but in your case this work-around may just do the job.. Use reference model restraints for the misbehaving part of the model. You can do that in two ways: 1) by restraining to the initial model itself or 2) by providing an external model for that region. I'm not sure which one would work best in your case soyou may end up trying both. Let me know if you need any help with this and I will help you set this up (off-list). For this I may need your model and map files so that I send you back a working example. Pavel On 5/15/23 07:03, CUNIASSE Philippe wrote:

Dear all,

I am trying to do a phenix.real_space_refine in a CryoEM map at 4.2 Å with a model obtained by MDFF (NAMD, a protein DNA complex) because the resulting structure, despite a good CC and a good molecular geometry has a poor Ramachandran Map.

When doing the phenix.real_space_refine, the CC is slightly increased with respect to the starting structure, the Ramachandran is dramatically increased but some part of the model ( in the case, the first three bases pairs a DNA duplex) are strongly distorted because of a lack of density in the corresponding region of the map.

For some reason, this does not append with the MDFF method.

I was wondering if there is a mean, in phenix.real_space_refine, to exclude a part of the model from being fitted in the map. Indeed, the distortion I observe in the DNA duplex is likely due to the fact that the real_space_refine tries to put the atoms at the extremity of the duplex in the closest density at the price of a strong distortion of the structure.

Thank in advance for your suggestions.

Best regards.

Philippe.